Abstract Introduction Due to its peculiar anatomy and physiology, the pericardial fluid is a biological matrix of particular interest in the forensic field. Despite this, the available literature has mainly focused on post-mortem biochemistry and forensic toxicology, while to the best of authors’ knowledge post-mortem metabolomics has never been applied. Similarly, estimation of the time since death or post-mortem interval based on pericardial fluids has still rarely been attempted. Objectives We applied a metabolomic approach based on ¹H nuclear magnetic resonance spectroscopy to ascertain the feasibility of monitoring post-mortem metabolite changes on human pericardial fluids with the aim of building a multivariate regression model for post-mortem interval estimation. Methods Pericardial fluid samples were collected in 24 consecutive judicial autopsies, in a time frame ranging from 16 to 170 h after death. The only exclusion criterion was the quantitative and/or qualitative alteration of the sample. Two different extraction protocols were applied for low molecular weight metabolites selection, namely ultrafiltration and liquid-liquid extraction. Our metabolomic approach was based on the use of ¹H nuclear magnetic resonance and multivariate statistical data analysis. Results The pericardial fluid samples treated with the two experimental protocols did not show significant differences in the distribution of the metabolites detected. A post-mortem interval estimation model based on 18 pericardial fluid samples was validated with an independent set of 6 samples, giving a prediction error of 33–34 h depending on the experimental protocol used. By narrowing the window to post-mortem intervals below 100 h, the prediction power of the model was significantly improved with an error of 13–15 h depending on the extraction protocol. Choline, glycine, ethanolamine, and hypoxanthine were the most relevant metabolites in the prediction model. Conclusion The present study, although preliminary, shows that PF samples collected from a real forensic scenario represent a biofluid of interest for post-mortem metabolomics, with particular regard to the estimation of the time since death.